Table 10.

Comparison of different variants of our method on Caltech101. The best performance indicators of all the taxonomies are marked as bold.

	ori.	cri. 1	cri. 2	cri. 3	cri. 4	cri. 5	$\overline{\mathit{\xi }}$ ± s
with all components	0.50	0.54	0.59	0.56	0.54	0.59	0.55 ± 0.03
No log-ratio loss	0.45	0.43	0.47	0.43	0.43	0.43	0.44 ± 0.01
No feature learning	0.38	0.38	0.36	0.34	0.37	0.37	0.37 ± 0.01
No active selection	0.49	0.52	0.51	0.53	0.51	0.58	0.52 ± 0.03
No verification	0.46	0.47	0.46	0.47	0.46	0.48	0.47 ± 0.01
No weighted blending	0.38	0.49	0.40	0.43	0.44	0.53	0.45 ± 0.05
No CRF	0.49	0.53	0.53	0.53	0.52	0.56	0.53 ± 0.02
No verification stopping	0.26	0.24	0.32	0.32	0.29	0.40	0.30 ± 0.05